An Nvidia-backed AI startup is planning a multi-billion-dollar data centre in South Korea, positioning it as a bulwark against the adoption of Chinese open-source AI in the region.
The data centre will power open-source AI models developed by Reflection AI, a startup founded two years ago by two former DeepMind researchers. Models trained in the facility will be tuned to the Korean language and culture, potentially opening the door to more locally trained and marketed AI apps and services.
It is a tall task for a company that has yet to publish an AI model of its own, instead relying on the reputation of its two co-founders and their previous work on AlphaGo, PaLM, GPT-4, and Google Gemini. The company has already raised over $2 billion from Nvidia and an investment group backed by Donald Trump Jr, according to the Wall Street Journal.
The open-source models would most likely follow a similar design to popular Chinese AI models such as DeepSeek and Alibaba’s Qwen. These are not truly open source, but offer greater code transparency and far more scope for local customisation than US models like GPT and Claude.
Nvidia will supply thousands of chips for the data centre, another sign of the circular nature of its investments in AI companies. In November, it announced plans to invest $10 billion in the country, supplying over a quarter of a million GPUs to firms.
The data centre will be built in partnership with retail conglomerate Shinsegae Group, which will provide financing, real estate, and permits. The group also has ties to Donald Trump Jr., and the arrangement appears to align with the US administration’s goal of improving competitiveness in the AI market.
Nvidia funding the expansion of AI infrastructure
Nvidia has been at the centre of the surge in AI infrastructure spending over the past 12 months. As the leader in the GPU market, it has ramped up production of its most powerful GPU servers for AI training and inference. This pushed total revenues in 2025 to $215 billion, with more than half reported as profit.
With this level of cash generation, Nvidia has become one of the most active investors in both AI companies and infrastructure buildouts. It invested $30 billion into OpenAI in its latest funding round, one of the largest single investments into a private company. It is also involved in OpenAI’s $500 billion Stargate data centre programme as a key supplier of GPUs and infrastructure.
Outside of OpenAI, Nvidia was the lead investor in a $2 billion funding round for Nscale, an AI data centre startup building specialised facilities for AI workloads. It has also partnered with Cisco to develop next-generation data centres.
It has also invested $5 billion in rival Intel. With Intel’s growth stalling in the AI era, Nvidia increasingly appears to view it less as a competitor and more as a partner in scaling data centre infrastructure. It has taken a similar approach with memory chip manufacturers, which are under pressure from surging AI-driven demand.
Also read: Learn more about Nvidia’s AI infrastructure push in our report on Nscale’s $2 billion raise.
